OBJECTIVE: To determine the trends in incidence of diagnosis of bronchopulmonary dysplasia (BPD) and associated health services use for the neonatal hospitalization of patients with BPD in an era of changing definitions and management.
PATIENTS AND METHODS: All neonatal hospitalization records available through the Nationwide Inpatient Sample, 1993–2006, were analyzed. Multivariable regression analyses were performed for incidence of BPD diagnosis and associated hospital length of stay and charges. Multiple models were constructed to assess the roles of changes in diagnosis of very low birth weight (VLBW) neonates and different modalities of respiratory support used for treatment.
RESULTS: The absolute incidence of diagnosis of BPD fell 3.3% annually (P = .0009) between 1993 and 2006 coincident with a 3.5-fold increase in the use of noninvasive respiratory support in patients with BPD. When data were controlled for demographic factors, this significant decrease in incidence persisted at a rate of 4.3% annually (P = .0002). All models demonstrated a rise in hospital length of stay and financial charges for the neonatal hospitalization of patients with BPD. The incidence of BPD adjusted for frequency of prolonged mechanical ventilation also decreased but only by 2.8% annually (P = .0075).
CONCLUSIONS: The incidence of diagnosis of BPD decreased significantly between 1993 and 2006. In well-controlled models, birth hospitalization charges for these patients rose during the same period. Less invasive ventilatory support may improve respiratory outcomes of VLBW neonates.
WHAT'S KNOWN ON THIS SUBJECT:
There is discrepancy in the literature about whether the incidence of BPD is stable or decreasing as the care of premature neonates improves. Little is known about health care utilization during neonatal hospitalization for infants with BPD.
WHAT THIS STUDY ADDS:
This study is an evaluation of the incidence and associated hospital charges and LOS for neonates diagnosed with BPD between 1993 and 2006. To the authors knowledge, no other nationwide analysis of this type has been reported.
Bronchopulmonary dysplasia (BPD) is a chronic lung disease that stems from premature birth. Estimates of the incidence of BPD based on multicenter studies at academic medical centers indicate that BPD is diagnosed at the time of discharge from the NICU in 25% to 35% of very low birth weight (VLBW) infants (those born with a birth weight of <1500 g).1 Although BPD remains a significant morbidity associated with preterm birth, measurements of the incidence of this disease, and in particular of trends in the incidence, remain elusive. Obtaining a better understanding of changes in BPD incidence over time is tremendously important, because increasing numbers of neonatal patients are at risk for the disease, because advances in fertility treatments and neonatal care are leading to the survival of greater numbers of premature neonates with viability periods earlier in gestation. Elucidation of the financial impact of BPD on the cost of neonatal care is equally important in the current climate of rising intensive care costs and shrinking health care resources.
In 1967, Northway et al2 defined BPD as a lung disease of premature infants who required supplemental oxygen at 28 days of life. The patients identified by Northway et al were born after ∼32 weeks' gestation. As survival of extremely premature infants improved, it became necessary to alter the original definition of BPD. Since the late 1990s, BPD has been generally defined by the need for supplemental oxygen at 36 weeks' postconceptional age.3 In recent years BPD has been diagnosed in a patient population different from that seen by Northway et al in the 1960s. Infants with the “new” BPD4,5 are born much earlier in gestation and have not reached term-equivalent development by 1 month of age. In these neonates the disease is attributed predominantly to abnormal pulmonary development, because surfactant deficiency is now treatable and mechanical ventilators are less damaging.
Although studies have evaluated the impact of new pharmacologic and respiratory therapies on BPD outcomes, these investigations have generally been limited to tertiary care centers.6,–,9 To our knowledge, no study has examined the nationwide impact of changes in BPD management across all hospital types. The Nationwide Inpatient Sample (NIS) is a valuable tool for addressing questions about the use of health services at the national level. The NIS is a statistically representative sample of the full diversity of hospital care in the United States during the past 2 decades. From its inception in 1988 as a small database of discharge information on hospital patients, the NIS has grown to become a statistically accurate representation of inpatient care in the United States. The NIS currently includes discharge data from 1054 hospitals in 37 states, which represents a 20% stratified sample of hospitals in the United States. In this ecological study, we used nationally representative hospital-discharge data from a 14-year period to examine the impact of changes in management on the incidence of BPD, as well as associated length of stay (LOS) and hospital charges for patients with BPD during the neonatal hospitalization.
We analyzed a multiyear data file concatenated from the 1993–2006 NIS, part of the Health Care Cost and Utilization Project (HCUP) of the Agency for Healthcare Quality and Research.10 A stratified sample of approximately one-fifth of US hospitals is selected each year for the NIS. Discharge-level weights are provided to facilitate calculation of national estimates for discharges. Although available, the 1988–1992 data sets included a smaller, less-representative subset of US states than later years. We evaluated data available for neonates from this time period and identified inconsistencies likely attributable to data-collection methods. For this reason, we eliminated the years 1988–1992 from our analysis, according to the recommendations of the HCUP, which were documented in a white paper regarding interpretation of multiyear trends.11 We proceeded to analyze the 1993–2006 data set and refer to those years as the “entire” data set throughout this report.
Because data were collected and deidentified before this study, the institutional review board of the Mount Sinai School of Medicine deemed this project exempt from review.
Hospital discharge diagnoses and procedures were coded in accordance with the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM). Hospitalizations of interest were narrowed to the neonatal hospitalization by using the HCUP codebook variable NEOMAT. Among these neonatal hospitalizations, patients with BPD were identified by use of the ICD-9-CM patient diagnosis code 770.7 (chronic respiratory disease arising in perinatal period, including BPD). Within this group of hospitalizations, diagnosis codes of interest were investigated by using diagnosis codes for neonates with appropriate growth with birth weight at <1500 g (764.01–764.05, 764.11–764.15, 764.21–764.25, 764.91–764.95, 765.01–765.02, 765.11–765.15). Procedures of interest were also investigated by using procedure codes 93.90 (noninvasive mechanical ventilation without endotracheal tube or tracheostomy) and 96.72 (continuous invasive mechanical ventilation for ≥96 hours).
Descriptive data, including LOS, hospital charges, demographic characteristics (race, gender, median income quartile for patient zip code, primary payer), hospital characteristics (region of country, rural/urban, teaching status), procedural and diagnostic frequency, and frequency of death during hospitalization were identified by using the appropriate HCUP codebook variables and presented in comparison to all other neonatal hospitalizations. Univariate regression analysis was performed for incidence, LOS, and financial charges with each of the demographic, hospital characteristic, and procedural independent variables. Because of their skewed distribution, log (base 10) transformation of LOS and charges were used in all univariate and multivariable analyses. All models were controlled for inflation by using the Medical Care Consumer Price Index.12
To assess (1) the change in incidence of diagnosis of BPD among neonates, (2) the birth-hospitalization length for neonates with BPD, and (3) the birth-hospitalization charges for neonates with BPD, we performed multivariable analyses with these 3 primary outcomes as dependent variables and used the period 1993–2006 and all the above-listed hospital and demographic variables as independent variables. We then endeavored to additionally control for changes in the incidence of low birth weight by adding the incidence of VLBW births as an independent variable. Finally, we examined the implications of adding the use of noninvasive respiratory support, and prolonged (≥96 hours) intubation and mechanical ventilation in 2 separate sets of regressions to examine the impact of each of these independent variables. We analyzed all methods of noninvasive respiratory support together (including continuous positive end-expiratory pressure, nasal synchronized intermittent mandatory ventilation, and high-flow nasal cannula), but for simplicity we referred to this category as CPAP (continuous positive airway pressure). Statistical analyses were conducted by using SAS 9.1.3 (SAS Institute Inc, Cary, NC) and SUDAAN 9.0.3 (Research Triangle Institute, Research Triangle Park, NC).
We included data from 9 542 032 neonatal hospitalizations in our analysis, which represents a weighted statistical sample of 47 465 901 neonatal hospitalizations.
Demographic information for neonatal hospitalizations examined in the NIS is listed in Table 1. There was significant variability for hospital charges and LOS according to demographic and hospital characteristics, results that supported the use of controls for these variables in multivariable models. In particular, BPD was diagnosed in a disproportionate percentage of neonates with Medicaid insurance and those in the lowest median income quartile for home zip code. In addition, BPD was diagnosed in a disproportionate percentage of neonatal patients who were identified as black and in a disproportionate percentage of neonatal patients cared for at nonteaching hospitals.
Trends in diagnoses and procedures of interest are depicted in Fig 1. Between 1993 and 2006, the incidence of BPD diagnosis fell 3.3% annually (P = .0009). Coincidentally, the use of CPAP in neonates increased 8.54% annually (P < .0001). The increase in use of CPAP among neonates diagnosed with BPD was even more striking, because it rose from 6.4% in 1993 to 22.6% in 2006. The incidence of the use of prolonged (≥96 hours) mechanical ventilation in neonates did not change significantly between 1993 and 2006 (P = .85), nor did the incidence of diagnosis of VLBW (P = .25).
Our multivariable analyses resulted in 4 distinct sets of models to evaluate trends in health resource use for neonates with BPD. The first set of models was used to investigate trends in the incidence of neonatal diagnosis of BPD and associated hospital LOS and charges over time. The second set of models was used to evaluate trends in neonatal BPD incidence, LOS, and charges. These models were controlled for changes in the incidence of VLBW in neonatal patients. Additional models were used to evaluate BPD incidence, LOS, and charges and were controlled for both the incidence of VLBW in neonates and the use of CPAP or prolonged mechanical ventilation (variable VENT).
Model 1: Incidence of BPD and Trends in Hospital LOS and Charges
When data were controlled for demographic factors, there was an overall 4.3% annual decrease (P = .0002) in diagnosis of BPD among neonates who survived to discharge during the period from 1993 to 2006 (Fig 2A). Despite this decline in incidence of BPD, both hospital LOS and charges increased significantly between 1993 and 2006 for neonates with BPD, with annual increases of 3.9% and 4.9%, respectively (each P < .0001).
Model 2: Incidence of BPD and Trends in Hospital LOS and Charges Controlled for VLBW
When data were controlled for demographic factors and for the slight increase in VLBW births seen between the early and later years of the NIS, the decrease in incidence of BPD over the entire data set persisted at a rate of 5.0% annually (P < .0001) (Fig 2A). The annual increase in hospital LOS and charges over the entire data set also persisted, at annual rates of increase of 2.0% and 2.9%, respectively (each P < .0001).
Model 3: Incidence of BPD and Trends in Hospital LOS and Charges Controlled for VLBW and CPAP
When use of CPAP was added to the model, the annual incidence of BPD over the entire data set decreased by 5.5% annually (P < .0001) from 1993 to 2006 (Fig 2B). Although increases were smaller than in the models for which CPAP was not considered, annual hospital LOS and total charges rose in this model by 1.8% and 2.7%, respectively (each P < .0001).
Model 4: Incidence of BPD and Trends in Hospital LOS and Charges Controlled for VLBW and VENT
Our final model was used to evaluate BPD and controlled for use of long-term invasive mechanical ventilation as well as demographic data and VLBW. In this model the incidence of BPD decreased by 2.8% annually (P = .0075), whereas LOS rose 2.1% annually (P < .0001) and charges rose 3.1% annually (P < .0001) (Fig 2B).
The major findings that resulted from this analysis were that during the past 14 years significant decreases in the incidence of diagnosed BPD have occurred and that birth-hospitalization charges and LOS for patients with BPD have increased significantly, according to models that controlled for increases in the incidence of VLBW.
An additional implication of our study results is that modern strategies of respiratory support have likely improved long-term respiratory outcomes in infants with VLBW. We evaluated the impact of invasive and noninvasive respiratory support on incidence of BPD and hospital use for these patients. The use of noninvasive respiratory support has been advocated since the late 1980s as a management strategy to decrease the incidence of chronic lung disease of prematurity.13 Our results support this recommendation. That said, prolonged mechanical ventilation with modern respirators might not be as harmful as previously thought. Our analysis results indicate that the incidence of diagnosis of BPD in VLBW infants decreased, if at an attenuated rate, when prolonged mechanical ventilation (mechanical ventilation for ≥96 hours) was added to the model. This finding should be explored further in patient-oriented studies.
A major benefit of the approach taken in this study is the use of a nationally representative sample in endeavoring to examine ongoing questions about how BPD management has affected outcomes over time. Our analysis data represent a weighted statistical sample of 47 465 901 neonatal hospitalizations. National health statistics from this time period indicate that 52 038 794 live births actually occurred between January 1, 1993, and December 31, 2006.14 The incidence of BPD seen in our study is comparable to that documented in prospective studies. For example, in 2004 Walsh et al reported an incidence of BPD ranging from 25% to 35%. The national incidence of BPD in the NIS in 2002 was 33.7%, which is in good agreement with this published incidence, which was determined in a prospective manner.
There is discrepancy in the literature as to whether the incidence of BPD in premature infants is decreasing because of improvements in prenatal care of high-risk pregnancies and intensive care of neonates or remaining stable despite improvements in other pediatric outcomes after premature birth. Results of several small, single-center studies15,16 indicate that the incidence of BPD has fallen during recent years. Results of other larger but older studies17 have indicated that the incidence of BPD remains unchanged. Although all longitudinal studies are challenged by the changing face of the patient population at risk for BPD, our analysis includes the demonstrated large percentage of patients with BPD who were cared for outside of academic medical centers. Given that previous evaluations of BPD outcomes were conducted almost exclusively at academic medical centers, our analysis provides insight into the care of a large, previously unstudied population.
We used birth weight as a surrogate for gestational maturity in our analysis, because discharge diagnoses are more commonly coded by birth weight than gestational age because of the objectivity of this measure. In the analyses adjusted for VLBW, we included birth-weight codes that represented appropriately grown neonates as opposed to the separate birth-weight codes for individuals with poor intrauterine growth, in an attempt to exclude small but mature neonates from our analysis. We also limited our analysis to neonates with birth weight of <1500 g, rather than performing analyses based on the more inclusive definition of low birth weight (<2000 g), in an additional attempt to limit our analysis to the more immature neonates most significantly at risk for chronic lung disease.
Much discussion in the neonatal intensive care literature suggests that avoiding the use of mechanical ventilation or minimizing its use outside of the first hours of life prevents or reduces the development of BPD. Our use of the ICD-9 code 96.72, continuous invasive mechanical ventilation for ≥96 hours, addresses this suggested change in management. Although most VLBW neonates require CPAP at some point during their birth hospitalization because of pulmonary immaturity (independent of chronic lung disease), those managed under modern, less invasive strategies will not receive mechanical ventilation for more than the initial hours of life during stabilization and administration of exogenous surfactant. Our CPAP variable did not exclude those infants who received invasive mechanical ventilation, but our VENT variable specifically addressed those who received significant mechanical ventilation.
Because a gradual change in the definition of BPD has been effected over the past 20 years and a detailed nationally accepted standard definition of BPD does not currently exist, our analysis cannot fully account for artifacts related to the changes in the definition of the disease that occurred during the late 1990s to early 2000s. Figure 1 shows a slight decrease in the incidence of diagnosis of BPD between 1999 and 2001, possibly related to the National Institutes of Health consensus statement change in definition of BPD that occurred in 2000.18 This definitional change was necessary because independent pulmonary function cannot be expected to occur in extremely preterm neonates at 28 days of life, given that many such infants have not reached even the late-preterm period of development 1 month after birth. That said, different NICUs adopted a variety of subtly different definitions of BPD at various time points in the late 1990s and early 2000s. It is impossible for any analysis to fully control for this evolution in the definition of BPD, and it is entirely possible that a portion of the decrease in incidence in BPD elucidated in this report can be accounted for by the changes in definition that occurred during this period of time.
With this investigation, as with any ecological study, caveats must be applied before conclusions can be drawn from correlations suggested by the results. In addition, important limitations exist that are attributable to the quality of available data and the vagaries of changes in standards of care management and uses of diagnostic codes in medical practice. Discharge diagnosis codes are sometimes incomplete or inaccurate. The large data set provided by the NIS, however, and the national sampling strategy used to accumulate these data mitigate this issue.
An additional problem arises with the use of discharge information to evaluate neonatal disease and demographic data, particularly hospital LOS and charges associated with neonatal hospitalization. As the model of organized neonatal care has evolved since its inception in the 1960s, transport of ill neonates from the hospital of delivery to tertiary care centers has been a mainstay of care. Although women known to have high-risk pregnancies are increasingly transferred to tertiary care centers before delivery, a significant number of extremely premature or newborns otherwise at high risk are still delivered at hospitals without facilities to care for them. Organized referral networks exist nationwide to facilitate rapid transfer of these patients to level 3 nurseries. For many years of NIS data, it is difficult or impossible to discern whether a patient discharge resulted in admission to another acute care hospital (either initial transfer or back transfer) or final discharge to home or to a chronic care facility. Therefore, some hospitalizations included in this analysis represent only the portion of a patient's care performed before or after transfer. Because of the balanced sampling strategy of the NIS, however, the data set should not contain an excess of data for partial hospitalization periods that occurred before or after transfer. Because of its exceedingly large size, the sample should provide a statistically representative picture of health care use of neonates with BPD on aggregate. Despite this peculiarity of neonatal care, discharge diagnosis codes have been previously validated for study of NICU outcomes and health care use.19,–,21
In this study we investigated the effects of changes in practice management of BPD as reflected in a national data set covering a 14-year period. On the basis of our results for data analyses controlled for the increase in VLBW births, we conclude that the incidence of BPD has decreased since the early 1990s after multiple changes in management strategies aimed at improving long-term pulmonary outcomes. Global improvements in neonatal care have clearly had an impact on functional pulmonary outcomes in neonates.
We thank Dr Ian Holzman for helpful suggestions and support of this project.
- Accepted April 21, 2010.
- Address correspondence to Annemarie Stroustrup, MD, MPH, Division of Newborn Medicine, Kravis Children's Hospital, Mount Sinai Medical Center, One Gustave L. Levy Place, Box 1508, New York, NY 10029. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
- BPD =
- bronchopulmonary dysplasia •
- VLBW =
- very low birth weight •
- NIS =
- Nationwide Inpatient Sample •
- LOS =
- length of stay •
- HCUP =
- Health Care Cost and Utilization Project •
- ICD-9-CM =
- International Classification of Diseases, Ninth Revision, Clinical Modification •
- CPAP =
- continuous positive airway pressure
- Walsh MC,
- Yao Q,
- Gettner P,
- National Institute of Child Health and Human Development Neonatal Research Network
- Jobe AH
- Walsh MC,
- Yao Q,
- Horbar JD,
- Carpenter JH,
- Lee SK,
- Ohlsson A
- Stoelhorst GM,
- Rijken M,
- Martens SE,
- Leiden Follow-Up Project on Prematurity
- 10.↵Healthcare Cost and Utilization Project. HCUP Databases: Overview of the Nationwide Inpatient Sample (NIS). Rockville, MD: Agency for Healthcare Research and Quality; 2008. Available at: www.hcup-us.ahrq.gov/nisoverview.jsp. Accessed October 30, 2008
- 11.↵Healthcare Cost and Utilization Project. User Guide for the 1988–2002 NIS Trends Supplemental Files. Rockville, MD: Agency for Healthcare Research and Quality; 2006. Available at: www.hcup-us.ahrq.gov/db/nation/nis/UserGuideNISTrendsSupplementalFilesv8.pdf. Accessed June 3, 2010
- 12.↵US Department of Labor, Bureau of Labor Statistics. Consumer Price Index. U.S. Medical Care, 1982–84=100 - CUUR0000SAM. Washington, DC: US Department of Labor. Available at: http://data.bls.gov/cgi-bin/surveymost?cu. Accessed February 8, 2008
- Avery ME,
- Tooley WH,
- Keller JB,
- et al
- 14.↵Centers for Diseases Control and Prevention. National Vital Statistics System. Atlanta, GA: Centers for Diseases Control and Prevention. Available at: www.cdc.gov/nchs/births.htm. Accessed February 2, 2010
- Geary C,
- Caskey M,
- Fonseca R,
- Malloy M
- 18.↵NIH Consens Statement. Antenatal corticosteroids revisited: repeat courses. 2000;17 (2):1–18
- Copyright © 2010 by the American Academy of Pediatrics